Struct StreamingOptions 

pub struct StreamingOptions {
    pub initial_batch_size: usize,
    pub throughput_batch_size: usize,
    pub fast_first_batch: bool,
    pub include_properties: bool,
    pub include_presentation_layers: bool,
    pub emit_quick_metadata_bootstrap: bool,
    pub retain_emitted_meshes: bool,
    pub tessellation_quality: TessellationQuality,
    pub entity_index: Option<Arc<EntityIndex>>,
    pub cancel: Option<Arc<AtomicBool>>,
    pub enable_instancing: bool,
}

Controls the tradeoff between first-frame latency and richer upfront metadata.

Fields

initial_batch_size: usize

Batch size used for the very first emitted chunk.

throughput_batch_size: usize

Batch size used after the first emitted chunk for higher throughput.

fast_first_batch: bool

Prioritize cheap/high-yield element classes first.

include_properties: bool

Include expensive property parsing on the first-frame path.

include_presentation_layers: bool

Include expensive presentation-layer resolution on the first-frame path.

emit_quick_metadata_bootstrap: bool

Emit a lightweight spatial bootstrap during the scan phase.

retain_emitted_meshes: bool

Retain emitted meshes in the returned ProcessingResult.

tessellation_quality: TessellationQuality

Tessellation detail level (#976). Medium reproduces the historical output byte-for-byte; consumer-selectable on the wasm path via setTessellationQuality, and on the server via the tessellation_quality query parameter. 2D symbolic extraction (symbolic.rs) deliberately ignores the level — symbols are resolution-independent line work.

entity_index: Option<Arc<EntityIndex>>

Pre-built entity index to reuse instead of scanning content again. A caller that runs both the geometry pass and a second pass over the same bytes (e.g. a server emitting GLB and extracting properties) can build_entity_index once and inject it here, skipping the duplicate SIMD scan. None builds it internally, exactly as before. The index MUST come from build_entity_index(content) for the same content, or decoding will read the wrong byte ranges.

cancel: Option<Arc<AtomicBool>>

Cooperative cancellation: when the flag flips true, the streaming core stops between job chunks (no further meshing or batch emission). The returned ProcessingResult is then PARTIAL - the caller set the flag, so it must not present the result as a completed parse. Used by the server to stop burning a core when an SSE client disconnects.

enable_instancing: bool

#1623 Phase 2 “don’t-bake” instancing. When true (AND retain_emitted_meshes), the geometry phase meshes each repeated single-solid IfcRepresentationMap ONCE (a template occurrence) and emits every OTHER occurrence as a lightweight InstanceRecord (placement + colour + id) instead of materializing a full world-space mesh per occurrence — killing the per-occurrence 43M-vertex bake on mapped-item-heavy models. false (default) reproduces the historical materialized output byte-for-byte, so determinism / parity / every exporter are unaffected (none arm this).

Trait Implementations

impl Clone for StreamingOptions

fn clone(&self) -> StreamingOptions

Returns a duplicate of the value.

fn clone_from(&mut self, source: &Self)

Performs copy-assignment from source.

impl Debug for StreamingOptions

fn fmt(&self, f: &mut Formatter<'_>) -> Result

Formats the value using the given formatter.

impl Default for StreamingOptions

fn default() -> Self

Returns the “default value” for a type.